JP2017147535A - White balance adjustment device, white balance adjustment method, white balance adjustment program, and photographing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a white balance adjustment device, a white balance adjustment method, a white balance adjustment program, and a photographing device that are suitable for white balance adjustment in which individual differences in skin color of a person is taken into consideration without increasing a processing load.SOLUTION: Normal white balance coefficient calculation processing S101 comprises: extracting a region estimated to be in an achromatic color from a subject image on the basis of an imaging signal; and using the achromatic color subject image to calculate a normal white balance coefficient. Skin color white balance coefficient calculation processing S102 comprises: extracting a region of a skin color from the subject image; and using a subject image of the skin color to calculate a white balance coefficient. White balance coefficient composition processing S103 comprises: putting together the normal white balance coefficient and the skin color white balance coefficient; and calculating a composite white balance coefficient. White balance adjustment processing S104 comprises using the composite white balance coefficient to adjust white balance of the imaging signal.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a white balance adjustment device, a white balance adjustment method, a white balance adjustment program, and a photographing device having a white balance adjustment function.

  An imaging device such as a digital camera or an imaging device mounted on a portable terminal incorporates a white balance adjustment function for obtaining stable color reproduction during color imaging. In the white balance adjustment function, adjustment is performed so that the pixel signal values of R (Red), G (Green), and B (Blue) are aligned in a state where an achromatic subject is photographed.

  In the white balance adjustment process, the light source that illuminates the subject is estimated, and the pixel signal value is adjusted based on the estimation result. However, for example, when the person is a subject and the light source cannot be estimated correctly, a color cast may occur in the skin color of the person's face, and a shot image with an uncomfortable feeling may be obtained. Therefore, for example, Patent Document 1 and Patent Document 2 disclose a photographing apparatus that uses a skin-colored region such as a person's face in a photographed image and corrects the color.

  Japanese Patent Application Laid-Open No. 2005-228561 discloses a configuration in which a human face is detected from a subject image and white balance adjustment processing is performed so that the detected skin color of the face becomes a reference skin color.

  Japanese Patent Application Laid-Open No. 2004-26883 discloses a configuration in which a white balance adjustment process is performed by estimating a light source that illuminates a subject. In the imaging device of Patent Document 2, a human face is detected from a subject image, and a melanin pigment component, a hemoglobin pigment component, and a shadow component are extracted from pixel signals in a region corresponding to the human skin, and these components are extracted. To estimate the light source.

JP 2010-041622 A JP2015-144420A

  In the white balance adjustment processing described in Patent Document 1, the pixel signal value is corrected so that the detected skin color becomes the reference skin color regardless of the actual skin color of the person. For this reason, the actual skin color of the person may not be suitably reproduced in the captured image.

  In Patent Document 2, a melanin pigment component, a hemoglobin pigment component, and a shadow component are extracted from a pixel signal. The pixel signal having the smallest melanin pigment component and hemoglobin pigment component is used for white balance adjustment processing, as it strongly reflects light source information. However, the subject image with the smallest pigment component when the subject image contains something other than human skin in the region used for white balance adjustment processing or when an error occurs in the face detection processing, There may occur a case where the subject image is other than human skin. In this case, since the white balance adjustment process is performed using a subject image other than the human skin, the light source may not be estimated correctly.

  Further, Patent Document 2 discloses a process for performing white balance adjustment processing using only a predetermined proportion of pixel signals from among the plurality of pixel signals having the lowest pigment component, instead of the pixel signal having the smallest pigment component. It is disclosed. However, in this processing, it is necessary to perform sorting (rearrangement) with a large processing load on a plurality of pixel signals based on the pigment component.

  The present invention has been made in view of the above circumstances, and its object is to perform white balance adjustment in consideration of individual differences in human skin color without increasing the processing load. To provide a white balance adjustment device, a white balance adjustment method, a white balance adjustment program, and a photographing device.

  According to one embodiment of the present embodiment, the white balance adjustment device detects an image signal storage unit that stores an image signal generated based on a light flux from a subject, and a skin area in the subject image based on the image signal. An independent component extraction unit that extracts an independent component from a color evaluation value of the skin region of the subject image, a statistic calculation unit that calculates a statistic that characterizes the distribution of the independent component, and a statistic A feature amount calculation unit that calculates a feature amount of the skin region based on the image data; and a first coefficient calculation unit that calculates a first coefficient used for white balance adjustment of the imaging signal using the feature amount.

  According to this embodiment, the white balance coefficient is calculated using the statistic of the independent component of the color evaluation value of the skin region. The statistic characterizes the distribution of a plurality of independent components, and is less influenced by some independent components. Therefore, by calculating the white balance coefficient using statistics, even when a subject image other than the human skin is included in a part of the skin region, the subject image other than the part of the skin is white. The influence on the balance coefficient can be suppressed. In addition, since the sorting of a plurality of independent components is not necessary for calculating the statistics of the independent components, according to the present embodiment, the skin color of a person in a captured image can be suitably reproduced while suppressing the processing load.

  Further, according to one embodiment of the present embodiment, for example, the feature amount calculation unit sets a reference value for the feature amount, detects a difference between the reference value and the feature amount, and according to the reliability of the statistic Change the calculation method of feature value and reference value. In this case, the first coefficient calculation unit calculates the first coefficient based on the difference detected by the feature amount calculation unit.

  Further, according to an embodiment of the present embodiment, the statistic is, for example, at least one of a distribution of a plurality of independent component distributions, a skewness of a plurality of independent component distributions, and a weighted average of a plurality of independent components. Including one.

  Further, according to one embodiment of the present embodiment, for example, the statistic calculation unit calculates a representative value of a plurality of independent component values, and the value is separated from the representative value by a predetermined value or more from the plurality of independent components. Statistics are calculated by excluding the independent components.

  Moreover, according to one Embodiment of this embodiment, an independent component contains a melanin pigment component and a hemoglobin pigment component, for example.

  Further, according to an embodiment of the present embodiment, the white balance adjustment mechanism calculates the second coefficient used for white balance adjustment of the imaging signal by a predetermined algorithm based on the color evaluation value of the imaging signal, for example. A second coefficient calculating unit; and a coefficient synthesizing unit that calculates a third coefficient used for white balance adjustment of the imaging signal based on the first coefficient and the second coefficient.

  Further, according to an embodiment of the present embodiment, the white balance adjustment mechanism further includes, for example, an achromatic color detection unit that extracts a region near the achromatic color in the subject image based on the imaging signal. In this case, the second coefficient calculation unit calculates the second coefficient based on the color evaluation value of the region near the achromatic color.

  According to an embodiment of the present embodiment, the white balance adjustment method includes an imaging signal storing step for storing an imaging signal generated based on a light flux from a subject, and detecting a skin area in the subject image based on the imaging signal. A skin detection step, an independent component extraction step for extracting an independent component from a color evaluation value of a skin area of the subject image, a statistic calculation step for calculating a statistic characterizing the distribution of the independent component, and a statistic A feature amount calculating step for calculating a feature amount of the skin region based on the image data; and a first coefficient calculating step for calculating a first coefficient used for white balance adjustment of the imaging signal using the feature amount.

  According to an embodiment of the present embodiment, the white balance adjustment program detects an imaging signal step for storing an imaging signal generated based on a light flux from a subject, and a skin area in the subject image based on the imaging signal. Based on the statistic, a skin detection step, an independent component extraction step for extracting an independent component from the color evaluation value of the skin area of the subject image, a statistic calculation step for calculating a statistic characterizing the distribution of the independent component A method including: a feature amount calculating step of calculating a feature amount of the skin region; and a first coefficient calculating step of calculating a first coefficient used for white balance adjustment of the imaging signal using the feature amount. This is a program to be executed.

  According to one embodiment of the present embodiment, an imaging device includes an imaging unit that generates an imaging signal generated based on a light flux from a subject, and a skin detection unit that detects a skin region in a subject image based on the imaging signal An independent component extraction unit that extracts an independent component from a color evaluation value of the skin region of the subject image, a statistic calculation unit that calculates a statistic that characterizes the distribution of the independent component, and a skin region based on the statistic And a first coefficient calculation unit that calculates a first coefficient used for white balance adjustment of the image pickup signal using the feature quantity.

  According to the present embodiment, a white balance adjustment device, a white balance adjustment method, a white balance adjustment program, and a white balance adjustment suitable for performing white balance adjustment in consideration of individual differences in human skin color without increasing processing load An imaging device is provided.

FIG. 1 is a block diagram illustrating a configuration of a photographing apparatus according to an embodiment of the present invention. FIG. 2 is a flowchart according to white balance adjustment processing according to the embodiment of the present invention. FIG. 3 is a flowchart according to the normal white balance adjustment processing in the embodiment of the present invention. FIG. 4 is a diagram showing a CxCyBv color space and a light source determination area in the embodiment of the present invention. FIG. 5 is a flowchart according to the flesh color white balance adjustment process in the embodiment of the present invention. FIG. 6 is an example of a subject image including a human face in the embodiment of the present invention. FIG. 7 is a diagram showing a CxCyBv color space and a skin color determination area in the embodiment of the present invention. FIG. 8 is a diagram showing a plane with the melanin pigment component and the hemoglobin pigment component as axes in the embodiment of the present invention. FIG. 9 is a diagram showing a plane with the melanin pigment component and the hemoglobin pigment component as axes in the embodiment of the present invention, and a graph showing the frequency of occurrence of each component. FIG. 10 is a diagram showing a plane with the melanin pigment component and hemoglobin pigment component as axes in the embodiment of the present invention, and a graph showing the frequency of occurrence of each component.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following, a digital single lens reflex camera will be described as an embodiment of the present invention. The embodiments of the present invention are not limited to digital single-lens reflex cameras, but include, for example, mirrorless single-lens cameras, compact digital cameras, camcorders, tablet terminals, PHS (Personal Handy phone System), smartphones, feature phones, portable game machines, and the like. The apparatus may be replaced with another type of apparatus having a white balance adjustment function.

[Configuration of the photographing apparatus 1]
FIG. 1 is a block diagram showing a configuration of the photographing apparatus 1 of the present embodiment. As shown in FIG. 1, the photographing apparatus 1 includes a system controller 100, an operation unit 102, a drive circuit 104, a photographing lens 106, a diaphragm 108, a shutter 110, an image sensor 112, an image sensor drive circuit 114, an image processing engine 116, A buffer memory 118, a card interface 120, an LCD (Liquid Crystal Display) control circuit 122, an LCD 124, and a ROM (Read Only Memory) 126 are provided.

  The operation unit 102 includes various switches necessary for the user to operate the photographing apparatus 1, such as a power switch, a release switch, and a photographing mode switch. When the user presses the power switch, power is supplied from the battery (not shown) to the various circuits of the photographing apparatus 1 through the power line. After supplying power, the system controller 100 accesses the ROM 126, reads out a control program, loads it into a work area (not shown), and executes the loaded control program to control the entire photographing apparatus 1.

  When the release switch is operated, the system controller 100 includes a drive circuit so that an appropriate exposure is obtained based on a photometric value measured by a TTL (Through The Lens) exposure meter (not shown) built in the photographing apparatus 1. The diaphragm 108 and the shutter 110 are driven and controlled via 104. More specifically, drive control of the aperture 108 and the shutter 110 is performed based on an AE function designated by a shooting mode switch, such as a program AE (Automatic Exposure), shutter speed priority AE, aperture priority AE, or the like. Further, the system controller 100 performs AF (Autofocus) control together with AE control. An active method, a phase difference detection method, a contrast detection method, or the like is applied to the AF control. Since the configuration and control of this type of AE and AF are well known, detailed description thereof is omitted here.

  The light flux from the subject passes through the photographing lens 106, the diaphragm 108, and the shutter 110 and is received by the image sensor 112. The image sensor 112 is, for example, a single-plate color CCD (Charge Coupled Device) image sensor having a Bayer-type pixel arrangement, and accumulates an optical image formed by each pixel on the light receiving surface as a charge corresponding to the amount of light. The image signal is converted into an image signal corresponding to each color of (Red), G (Green), and B (Blue) and output to the image sensor driving circuit 114. Note that the image sensor 112 may be another imager such as a CMOS (Complementary Metal Oxide Semiconductor) image sensor.

  The imaging element driving circuit 114 performs predetermined signal processing such as clamping and demosaicing on the imaging signal input from the imaging element 112 and outputs the processed signal to the image processing engine 116. The image processing engine 116 performs predetermined signal processing such as white balance adjustment and Y / C separation on the imaging signal input from the imaging element driving circuit 114 to generate a luminance signal Y and color difference signals Cb and Cr, The image is compressed in a predetermined format such as JPEG (Joint Photographic Experts Group). The buffer memory 118 is used as a temporary storage location for processing data when the image processing engine 116 executes processing.

  A memory card 200 is detachably inserted into a card slot of the card interface 120. The image processing engine 116 can communicate with the memory card 200 via the card interface 120. The image processing engine 116 stores the generated compressed image signal (captured image data) in the memory card 200 (or a built-in memory (not shown) provided in the image capturing apparatus 1).

  Further, the image processing engine 116 buffers the generated luminance signal Y and color difference signals Cb and Cr in a frame memory (not shown) in units of frames. The image processing engine 116 sweeps the buffered signal from each frame memory at a predetermined timing, converts it into a video signal of a predetermined format, and outputs it to the LCD control circuit 122. The LCD control circuit 122 modulates and controls the liquid crystal based on the image signal input from the image processing engine 116. Thereby, the photographed image of the subject is displayed on the display screen of the LCD 124. The user can view through the display screen of the LCD 124 a real-time through image captured with appropriate brightness and focus based on AE control and AF control.

  When the user performs a playback operation of the captured image, the image processing engine 116 reads the captured image data specified by the operation from the memory card 200 or the built-in memory, converts the image data into an image signal of a predetermined format, and the LCD control circuit 122. Output to. The LCD control circuit 122 performs modulation control on the liquid crystal based on the image signal input from the image processing engine 116, so that a captured image of the subject is displayed on the display screen of the LCD 124.

[White balance adjustment processing]
Next, signal processing performed on the signal input from the image sensor driving circuit 114 by the image processing engine 116 will be described. FIG. 2 shows a flowchart relating to white balance adjustment processing by the image processing engine 116. As shown in FIG. 2, in the white balance adjustment process, a normal white balance coefficient calculation process S101, a flesh color white balance coefficient calculation process S102, a white balance coefficient synthesis process S103, and a white balance adjustment process S104 are executed.

In the normal white balance coefficient calculation processing S101, an achromatic color is estimated from the subject image based on the imaging signals R _P , G _P , and B _P that are output from the imaging device 112 and subjected to signal processing by the imaging device driving circuit 114. Regions are extracted. Next, a known white balance coefficient (normal white balance coefficient) calculation process is performed using the extracted achromatic subject image. In the flesh color white balance coefficient calculation process S102, a flesh color region is extracted from the subject image, and a white balance coefficient (skin color white balance coefficient) calculation process is performed using the flesh color subject image. In the white balance coefficient combining process S103, the normal white balance coefficient and the skin color white balance coefficient are combined to calculate a combined white balance coefficient. In the white balance adjustment process S104, the white balance of the imaging signals R _P , G _P , and _BP is adjusted using the combined white balance coefficient.

[FIG. 3 (Normal White Balance Coefficient Calculation Processing)]
FIG. 3 shows a flowchart of the normal white balance coefficient calculation process S101.

[S201 in FIG. 3 (Image Division Processing)]
In this processing step S201, the subject image is divided into m (m is a positive integer) divided areas based on the imaging signal input from the imaging element driving circuit 114. Each divided region includes R, G, and B color information (color evaluation value). Based on the color information of each divided region, color signals R [i], G [i], and B [i] (i = 1 to m) corresponding to R, G, and B are calculated. The color signal is, for example, an average value of signal values representing the color information of each pixel in the divided area.

なお、輝度信号Ｙ[i]の算出において、各色信号Ｒ[i]、Ｇ[i]、Ｂ[i]にかけられる定数は数式１に示される値に限定されない。この定数は、ＬＣＤ１２４の仕様やビデオ信号の規格に応じて適宜変更可能である。 [S202 in FIG. 3 (Luminance Signal Generation Processing)]
In this processing step S202, a luminance signal Y [i] representing the luminance of each divided region is calculated based on the color signal of each divided region. The luminance signal Y [i] of the i-th divided area is, for example, ITU-R BT. This is a luminance signal of 601 standard and is calculated by the following formula 1.

In the calculation of the luminance signal Y [i], the constants applied to the color signals R [i], G [i], and B [i] are not limited to the values shown in Equation 1. This constant can be appropriately changed according to the specification of the LCD 124 and the standard of the video signal.

なお、Ｂｖ、Ｓｖ、Ｅｖは何れもＡＰＥＸ（Additive System of Photographic Exposure）値である。また、露出値Ｅｖは、絞り１０８の絞り値Ａｖと、シャッタ１１０のシャッタスピードＴｖの和で表されるため、数式２は次の数式３のように変形できる。

[S203 in FIG. 3 (Subject Luminance Calculation Unit)]
In this processing step S203, the subject brightness is calculated based on the brightness signal Y [i], the exposure value of the photographing apparatus 1 and the sensitivity of the image sensor 112. Assuming that the subject brightness when the subject is imaged with an appropriate exposure value is Bv, the sensitivity of the image sensor 112 is Sv, and the exposure value of the photographing device 1 is Ev, these relationships are expressed by the following Equation 2.

Bv, Sv, and Ev are all APEX (Additive System of Photographic Exposure) values. Further, since the exposure value Ev is expressed by the sum of the aperture value Av of the aperture 108 and the shutter speed Tv of the shutter 110, the equation 2 can be transformed into the following equation 3.

この輝度信号の差分Ｘｖ[i]を考慮すると、輝度信号Ｙ[i]によって補正された各分割領域の被写体輝度Ｂｖ[i]は、次の数式５のように表される。

なお、被写体輝度Ｂｖの算出方法は、上記方法に限定されない。例えば、被写体輝度Ｂｖの算出方法は、規格、撮影装置１の仕様、被写体の種類に応じて変更してもよい。また、被写体輝度Ｂｖの算出精度を向上するために、被写体像の周辺部に発生する撮像レンズ１０６によるビネッティング（ケラレ）が補正された上で被写体輝度Ｂｖが算出されてもよい。 Since Av, Tv, and Sv are known values obtained from the drive circuit 104 and the image sensor drive circuit 114, the subject brightness Bv can be calculated by Equation 3. However, generally, the subject is photographed at various brightness levels, and it is necessary to correct the subject brightness Bv according to the value of the brightness signal Y [i] calculated by Equation 1. For example, “CIPA DC-004-2004 Digital Camera Sensitivity Rules” established in July 2004 by the Camera and Imaging Products Association of Japan is known as the definition of appropriate exposure. In this sensitivity specification, the luminance signal Y = 118 is determined as appropriate exposure (standard output sensitivity) in an 8-bit system. Accordingly, if the luminance signal Y [i] when Y = 118 in the 8-bit system is set as Yb, the difference Xv [i] from the appropriate value of the luminance signal in each divided region is expressed by the following Equation 4. Is done.

Considering this luminance signal difference Xv [i], the subject luminance Bv [i] of each divided region corrected by the luminance signal Y [i] is expressed as in the following Equation 5.

Note that the method of calculating the subject brightness Bv is not limited to the above method. For example, the calculation method of the subject luminance Bv may be changed according to the standard, the specification of the photographing apparatus 1, and the type of the subject. Further, in order to improve the calculation accuracy of the subject brightness Bv, the subject brightness Bv may be calculated after correcting the vignetting (vignetting) caused by the imaging lens 106 generated in the peripheral portion of the subject image.

ここで、係数ｔ_x１、ｔ_x２、ｔ_ｙ１、ｔ_ｙ２は定数であり、後述する光源推定処理Ｓ２０５において使用される色空間に応じて適宜設定される。また、各色信号Ｃｘ[i]、Ｃｙ[i]は、色信号Ｒ[i]、Ｇ[i]、Ｂ[i]に所定の係数をかけて足し合わせることによって算出されてもよい。 [S204 in FIG. 3 (color signal generation processing)]
In this processing step S204, a color signal Cx [i] representing the color temperature of each divided area and a color signal Cy [i] representing the color deviation are calculated based on the color signal and luminance signal of each divided area. Cx [i] and Cy [i] are calculated by, for example, Expression 6 below.

Here, the coefficients t _x1 , t _x2 , t _y1 , and t _y2 are constants and are appropriately set according to the color space used in the light source estimation process S205 described later. Each color signal Cx [i], Cy [i] may be calculated by adding a predetermined coefficient to the color signals R [i], G [i], B [i].

[S205 in FIG. 3 (Light Source Estimation Processing)]
In this processing step S205, the color of the illumination light that illuminates the subject is estimated based on the subject brightness Bv [i] and the color signals Cx [i], Cy [i]. Specifically, in this processing step S205, each divided area has coordinates (Cx [i], Cy [i], Bv [i]) in the color space defined by the Cx axis, the Cy axis, and the Bv axis that are orthogonal to each other. Is plotted in Next, it is determined which of the plurality of light source determination areas defined in the color space is included in each plotted divided area.

  FIG. 4 shows a Cx-Cy plane when the color space is viewed along the Bv axis direction. The horizontal axis is the Cx axis representing the color temperature, and the vertical axis is the Cy axis representing the color deviation. The color space changes from red to blue as it goes from the negative direction of the Cx axis to the positive direction, and changes from green to magenta as it goes from the negative direction of the Cy axis to the positive direction. In FIG. 4, a curve indicated by a broken line is a black body radiation locus. A plurality of light source determination areas 301 to 307 are three-dimensionally defined in the color space. Each light source determination area corresponds to the color of illumination light that illuminates an achromatic subject. When the divided area is plotted in any one of the light source determination areas, the subject image in the divided area is regarded as an achromatic subject image illuminated by the corresponding color illumination light. . Therefore, it is possible to extract an achromatic region from the subject image and to estimate the light source of the illumination light by using the plurality of light source determination regions 301 to 307.

  Of the light source determination areas 301 to 307, the area 301 is an incandescent lamp color or light bulb color, the area 302 is warm white, the area 303 is sunlight color or white, the area 304 is cloudy color or day white, the area 305 Indicates shaded color or daylight color. A region 306 indicates a relatively reddish (low color temperature) fluorescent lamp color, and a region 307 indicates a relatively bluish (high color temperature) fluorescent lamp color.

  In this processing step S205, it is determined in which of the light source determination areas 301 to 307 each divided area is plotted. Next, the light source of the illumination light is estimated based on the plotted distribution of each divided region.

[S206 in FIG. 3 (coefficient determination processing)]
In this processing step S206, normal white balance coefficients G _Rw , G _Gw , and G _Bw are determined based on the estimated light source. These normal white balance coefficients are coefficients used in white balance adjustment processing S104 described later. Since the method for determining the white balance coefficient is generally well known, detailed description thereof is omitted here. If the white balance adjustment process S104 is performed using only the normal white balance coefficient, R _P , G _P , B _P for the imaging signal of each pixel, and R _P ′, G _P for the imaging signal after white balance adjustment. When placed as ′ and B _P ′, the adjusted imaging signals R _P ′, G _P ′, and B _P ′ are expressed by Equation 7 below.

When the normal white balance coefficients G _Rw , G _Gw , and G _Bw are determined in this processing step S206, the normal white balance coefficient calculation process S101 ends, and the next skin color white balance coefficient calculation process S102 is executed.

[FIG. 5 (Skin Color White Balance Coefficient Calculation Processing)]
FIG. 5 shows a flowchart of the flesh color white balance coefficient calculation process S102.

[S301 in FIG. 5 (Face Detection Processing)]
In this processing step S301, face detection processing in the subject image is performed based on the imaging signal input from the imaging element driving circuit 114. Since the algorithm of the face detection process is well known, detailed description thereof is omitted here. Further, the imaging signal subjected to the face detection process may be an imaging signal for main exposure stored as captured image data, or may be an imaging signal generated immediately before main exposure.

  Usually, the image capturing apparatus 1 continuously captures an image of a subject to determine exposure and composition before performing main exposure. For example, in a digital single-lens reflex camera, a subject is imaged by a dedicated sensor installed in the viewfinder in viewfinder shooting, and a subject is imaged by the same image sensor as in main exposure in live view shooting. Since the subject imaging process is performed until immediately before the main exposure is performed, if the time difference between the time at which the main exposure is performed and the time at which the immediately preceding imaging process is performed is sufficiently small, the subject between the time differences The movement or change is small. Therefore, by performing the face detection process using the imaging signal generated immediately before the main exposure, substantially the same face detection result as in the main exposure can be obtained. Further, in this case, the face detection process after the main exposure is not necessary, so that the processing load and processing time for the main exposure can be suppressed.

[S302 in FIG. 5 (region association processing)]
In this process step S302, the area (face area) in which the face in the subject image is detected by the face detection process is associated with the divided area divided by the image division process S201. Specifically, in this processing step S302, data (label) indicating that the face area is included is added to the divided area including the face area.

  FIG. 6 shows an example of a subject image including a human face. The subject image 600 is divided into a plurality of (m) divided regions by the image dividing process S201. When face detection processing S301 and region association processing S302 are performed on the subject image 600, for example, it is determined that the region 610 surrounded by the bold solid line in FIG. 6 is a face region. In this case, a label is added to the divided area included in area 610.

[S303 in FIG. 5 (skin color determination processing)]
In this processing step S303, it is determined whether or not each divided region to which a label has been added by the region association processing S302 is a skin color. Specifically, in this processing step S303, the subject luminance Bv [i] and the color signals Cx [i], Cy [i] of each divided area to which the label is added are a plurality of skin color determination areas defined in the color space. It is determined which of these is included.

  FIG. 7 shows a Cx-Cy plane when the color space is viewed along the Bv axis direction. The horizontal axis is the Cx axis representing the color temperature, and the vertical axis is the Cy axis representing the color deviation. In the color space, a plurality of skin color determination areas 401 to 404 are three-dimensionally defined. Each skin color determination region is defined for each color of illumination light that illuminates a skin color subject.

  Of the skin color determination regions 401 to 404, the region 401 corresponds to a skin color illuminated by a light source of any one of an incandescent lamp, a light bulb color, and warm white. A region 402 corresponds to a skin color illuminated by sunlight or a white light source. An area 403 corresponds to one of a skin color when cloudy, a skin color when the subject is in the shade, and a skin color illuminated with a daylight white or daylight light source. A region 404 corresponds to a skin color illuminated by light from a light source having poor color rendering properties such as a fluorescent lamp.

  When the divided area to which the label is added is plotted in any one of the skin color areas, the subject image of the divided area is regarded as the skin color subject image illuminated by the corresponding light source. . On the other hand, when the divided area to which the label is added is not included in any skin color determination area, the label of the divided area is deleted. With this process, the label of the divided area determined as the face area by the erroneous detection in the face detection process S301 is deleted.

  For example, in the subject image shown in FIG. 6, the divided area 610A is a subject image of a person's cheek. This divided area 610A is plotted in the skin color determination area 402 in the color space shown in FIG. Further, although the label is added to the divided area 610B by the area association process S302, the face of the person is not included. When the color in the divided area 610B is a color other than the skin color, for example, the skin color determination area is plotted outside the divided area 610B in FIG. In this case, the label of the divided area 610B is deleted by the skin color determination process S303.

[S304 in FIG. 5 (skin color search processing)]
In this processing step S304, a divided area corresponding to the skin color subject image is searched from the divided areas around the divided area to which the label is added. For example, there is a possibility that a subject image including the skin of a person such as a neck, a shoulder, and an arm exists around a divided region where a label is added and it is determined that a face region is included. In this processing step S304, the subject luminance Bv [i] and the color signals Cx [i] and Cy [i] of each divided area that does not have a label surrounding the divided area to which the label has been added are displayed in color. It is determined whether it is included in any of a plurality of skin color determination areas 401 to 404 defined in the space. Labels are added to the divided areas determined to be included in any of the skin color determination areas 401 to 404. As a result, the number of divided regions to which labels are added increases, and the flesh color white balance coefficient calculation accuracy can be improved.

  In the skin color search process S304, for example, in the subject image shown in FIG. 6, it is determined whether or not the divided areas in the area 620 surrounded by the broken thick line are included in the skin color determination areas 401 to 404. Of the divided areas in the area 620, for example, the divided area 620A is a subject image of a person's arm and is included in any of the plurality of skin color determination areas 401 to 404. Therefore, a label is added to the divided area 620A.

[S305 in FIG. 5 (independent component extraction processing)]
In this processing step S305, independent component extraction processing is performed on the color signals R [i], G [i], and B [i] of the divided areas to which the labels are added.

ここで、係数Ｋ（Ｋ_００〜Ｋ_１２）は定数であり、撮影装置１毎に予め設定されて撮影装置１内の記憶領域に記憶されている。 The layer structure of human skin can be modeled from the epidermis side into an epidermal layer containing melanin as a main pigment, a dermis layer containing hemoglobin as a main pigment, and a subcutaneous tissue. It has also been found that the color of human skin mainly depends on the pigment components of melanin and hemoglobin. Although the appearance of the skin color differs depending on the concentration of these pigments, the properties of human melanin and hemoglobin pigments themselves generally have no individual differences, and each pigment component can be handled independently. When an independent component corresponding to melanin in each divided region is M [i] and an independent component corresponding to hemoglobin is H [i], each independent component is calculated by the following Equation 8.

Here, the coefficient K (K _{00 to} K ₁₂ ) is a constant, and is preset for each photographing apparatus 1 and stored in a storage area in the photographing apparatus 1.

[S306 in FIG. 5 (abnormal value removal processing)]
In this processing step S306, the labels of the divided areas determined not to be flesh color are removed from the divided areas to which labels are added. Specifically, an upper limit value and a lower limit value are set for each of the independent components M [i] and H [i]. It is determined that the divided region where the value of the independent component exceeds the upper limit value or the divided region where the value of the independent component is lower than the lower limit value does not include the skin color subject image, and the label is removed.

  FIG. 8 is a plot of the divided regions to which labels have been added on a plane (MH plane) with the melanin pigment component and the hemoglobin pigment component as axes. The horizontal axis of FIG. 8 shows the magnitude of the independent component M [i] corresponding to melanin, and the vertical axis shows the magnitude of the independent component H [i] corresponding to hemoglobin. Since the divided areas to which the labels are added correspond to the subject determined to be flesh-colored, in FIG. 8, the independent components of each divided area are concentrated and plotted near their average values (avgM, avgH). ing. However, since each of the divided areas is an area including a plurality of pixels, each of the divided areas includes a subject other than the skin color, such as a divided area 610C including the eyes and a divided area 610D including the lips shown in FIG. May contain an image. As described above, the divided areas including the subject other than the skin color are plotted in an area away from the average value of the independent components as shown in FIG.

  In the abnormal value removal process S306, thresholds thM and thH are set for the independent components M [i] and H [i]. The threshold thM represents a difference from the average value avgM of the independent component M [i]. The threshold thH represents a difference from the average value avgH of the independent component H [i]. A divided region in which at least one of the independent components M [i] and H [i] is farther than the threshold value from the average value of the independent components is determined to be abnormal (not skin color), and the label is removed. As a result, it is possible to prevent a divided region including subjects other than the skin color such as eyes, lips, and hair from being used for determining the skin color white balance coefficient.

  In the abnormal value removal processing S306, the average value (avgM, avgH) of the independent components is used as a reference value, and a divided region where the independent component is separated from the reference value by a predetermined value or more is determined to be abnormal. The embodiment is not limited to this. For example, the reference value may be a value that represents an independent component of a plurality of divided regions, and may be a mode value or a median value of the independent component.

[S307 in FIG. 5 (coefficient determination processing)]
In this process step S307, the skin color white balance coefficients G _Rf , G _Gf , and G _Bf are determined. Specifically, in this processing step S307, first, a statistic characterizing the distribution of the independent components M [i] and H [i] of the divided region to which the label is added is calculated. Next, a skin color feature amount is calculated based on the statistical amount, and the skin color white balance coefficients G _Rf , G _Gf , and G _Bf are determined using the feature amount. By performing the white balance adjustment process using the skin color white balance coefficients G _Rf , G _Gf , and G _Bf , the photographed image can be appropriately adjusted even when there are individual differences in the person's skin color.

  As described above, the skin color of a person depends on the pigment components of melanin and hemoglobin. Therefore, the smaller the values of the two independent components M [i] and H [i], the smaller the influence of the pigment component on the color of the subject image. A subject image having small values of the independent components M [i] and H [i] is relatively greatly affected by light that illuminates the subject. Therefore, when determining the coefficient used for the white balance adjustment process, among the divided areas determined to be skin color, by increasing the contribution of the divided areas where the independent component values M [i] and H [i] are small. Even when there are individual differences in the skin color of a person, it is possible to perform an appropriate white balance adjustment in accordance with the color of the light source.

  FIG. 9 is a plot of the divided areas to which labels have been added after the abnormal value removal process S306 on the MH plane. The horizontal axis in FIG. 9 indicates the size of the independent component M [i] corresponding to melanin, and the vertical axis indicates the size of the independent component H [i] corresponding to hemoglobin. FIG. 9 shows a graph 9A showing the frequency of occurrence of the independent component M [i] and a graph 9B showing the frequency of occurrence of the independent component H [i]. The horizontal axis of the graph 9A indicates the value m [j] of the independent component M [i], and the vertical axis indicates the occurrence frequency frq_m of each value m [j]. Here, j is a positive integer. In the graph 9A, the value m [j] of the independent component M [i] is rounded at a predetermined interval, and can take P values. The horizontal axis of the graph 9B indicates the value h [k] of the independent component H [i], and the vertical axis indicates the occurrence frequency frq_h of each value h [k]. Here, k is a positive integer. In the graph 9B, the value h [k] of the independent component H [i] is rounded at a predetermined interval, and can take Q values. As shown in graphs 9A and 9B, each independent component has a distribution in which the vicinity of the average value has a peak.

ここで、avgＭは、ラベルが追加された複数の分割領域の独立成分Ｍ[i]の平均値である。devＭは、ラベルが追加された複数の分割領域の独立成分Ｍ[i]の分散である。skwＭは、ラベルが追加された複数の分割領域の独立成分Ｍ[i]の歪度である。avgＨは、ラベルが追加された複数の分割領域の独立成分Ｈ[i]の平均値である。devＨは、ラベルが追加された複数の分割領域の独立成分Ｈ[i]の分散である。skwＨは、ラベルが追加された複数の分割領域の独立成分Ｈ[i]の歪度である。ａ１〜ａ３、及び、ｂ１〜ｂ３は係数であり、予め設定された上で、撮影装置１内の記憶領域に記憶されている。 In the coefficient determination process S307, the skin color feature amounts biasM and biasH are calculated for each independent component based on the statistic characterizing the distribution of the independent components M [i] and H [i]. The feature amounts biasM and biasH are amounts representing deviations from the reference feature amounts biasM ₀ and biasH ₀ , and are determined so that the contribution of the divided region having a low independent component value becomes large. The feature amounts biasM and biasH are calculated by the following formula 9, for example.

Here, avgM is an average value of the independent components M [i] of a plurality of divided regions to which labels are added. devM is the variance of the independent component M [i] of the plurality of divided regions to which labels are added. skwM is the skewness of the independent component M [i] of a plurality of divided regions to which labels are added. avgH is an average value of the independent components H [i] of a plurality of divided regions to which labels are added. devH is the variance of the independent component H [i] of the plurality of divided regions to which labels are added. skwH is the skewness of the independent component H [i] of a plurality of divided regions to which labels are added. a1 to a3 and b1 to b3 are coefficients, which are set in advance and stored in a storage area in the photographing apparatus 1.

  As shown in Expression 9, the feature amount includes an average value of the values of the independent components and a variance indicating the spread of the distribution of the independent components. Therefore, all the divided areas to which labels are added can be reflected in the calculation of the feature amounts biasM and biasH. Further, for example, when there is a subject image close to the color of the skin even though it is not a person's skin, the abnormal value removal unit 36 may not be able to remove the label of the divided region including the subject image. However, by calculating the feature amounts biasM and biasH using the variances devM and devH, the influence of subject images other than the skin of a relatively small number (frequency) of the feature amounts biasM and biasH can be suppressed. .

  Further, as shown in Expression 9, the feature quantities biasM and biasH include skewness skewwM and skwH indicating the asymmetry of the distribution, respectively. The skewness skwM and skwH take positive values when the distribution peak shifts to the smaller independent component value and the skirt extends toward the larger value. On the other hand, the skewness skwM and skwH take negative values when the distribution peak shifts toward the larger independent component value and the tail extends toward the smaller value. Therefore, when the skewness skwM and skwH have positive values, the independent components are concentrated on a relatively small value and have a high correlation with the color of the light source. Therefore, by appropriately setting the coefficients a1 to a3 and b1 to b3, it is possible to increase the contribution of the divided regions whose independent component values are smaller than the average value to the feature amounts biasM and biasH. Thereby, it is possible to calculate the feature amounts biasM and biasH that have a high correlation with the color of the light source and suppress the influence of individual differences in human skin color.

ここで、wgt_m[j]、wgt_h[k]はそれぞれ、frq_m[j]、frq_h[k]に掛けられる係数であり、予め設定された上で撮影装置１内の記憶領域に記憶されている。 Note that the method of calculating the feature amounts biasM and biasH is not limited to Equation 9. For example, the feature amounts biasM and biasH may be calculated by a weighted average of independent components. In this case, the feature amounts biasM and biasH are calculated by the following Expression 10.

Here, wgt_m [j] and wgt_h [k] are coefficients multiplied by frq_m [j] and frq_h [k], respectively, which are set in advance and stored in a storage area in the photographing apparatus 1.

  In Equation 10, the feature quantities biasM and biasH are calculated by weighting the occurrence frequency of the independent components according to the values of the independent components M [j] and H [k]. Therefore, it is possible to reflect all the divided areas to which labels have been added in the calculation of the feature amounts biasM and biasH. Further, by increasing the coefficients wgt_m [j] and wgt_h [k] for the independent component having a small value, the feature amount biasM having a high correlation with the color of the light source and suppressing the influence of individual differences in the color of the human skin. biasH can be calculated.

In FIG. 9, the feature amounts calculated by Equation 9 or Equation 10 are plotted on the coordinates (biasM, biasH). Further, in FIG. 9, the reference feature quantity is plotted on the coordinates (biasM ₀ , biasH ₀ ). The reference feature amounts biasM ₀ and biasH ₀ are independent components of the skin-colored object illuminated by the reference light source, and are determined in advance and stored in the storage area of the photographing apparatus 1. Note that the reference feature amounts biasM ₀ and biasH ₀ are not fixed and may be changed. For example, the reference feature amounts biasM ₀ and biasH ₀ are changed depending on which of the plurality of skin color determination regions 401 to 404 shown in FIG. May be.

これにより、撮影された被写体像と基準となる肌色の被写体像との差分を表す色信号rstＲ、rstＧ、rstＢが求められる。次に、数式１２により、肌色ホワイトバランス係数Ｇ_Ｒｆ、Ｇ_Ｇｆ、Ｇ_Ｂｆが計算される。

In the coefficient determination process S307, when the feature amounts biasM and biasH are calculated, skin color white balance coefficients G _Rf , G _Gf , and G _Bf are calculated based on the calculated feature quantities biasM and biasH. Specifically, the coefficient determination unit 37 first converts the difference between the feature amounts of the independent components (biasM−biasM ₀ , biasH−biasH ₀ ) into the RGB space using the following formula 11.

As a result, color signals rstR, rstG, and rstB representing the difference between the photographed subject image and the reference skin color subject image are obtained. Next, the skin color white balance coefficients G _Rf , G _Gf , and G _Bf are calculated by Expression 12.

When the skin color white balance coefficients G _Rf , G _Gf , and G _Bf are determined by the coefficient determination process S307, the skin color white balance coefficient calculation process S102 ends, and the next white balance coefficient synthesis process S103 is executed.

数式１３は、通常ホワイトバランス係数Ｇ_Ｒｗ、Ｇ_Ｇｗ、Ｇ_Ｂｗを、肌色ホワイトバランス係数Ｇ_Ｒｆ、Ｇ_Ｇｆ、Ｇ_Ｂｆによって補正するものである。肌色ホワイトバランス係数Ｇ_Ｒｆ、Ｇ_Ｇｆ、Ｇ_Ｂｆによる補正の度合い（重み）は、定数αによって決まる。定数αは、固定値であってもよく、通常ホワイトバランス係数Ｇ_Ｒｗ、Ｇ_Ｇｗ、Ｇ_Ｂｗ、及び、肌色ホワイトバランス係数Ｇ_Ｒｆ、Ｇ_Ｇｆ、Ｇ_Ｂｆの信頼度に応じて変更可能であってもよい。 [S103 in FIG. 2 (White Balance Coefficient Synthesis Process)]
In this processing step S103, the normal white balance coefficients G _Rw , G _Gw , G _Bw calculated in the normal white balance coefficient calculation process S101, and the flesh color white balance coefficient G _Rf calculated in the flesh color white balance coefficient calculation process S102, G _Gf, based on _{G Bf,} synthetic white balance coefficient for use in the present exposure _{G R, G G,} is _{G B} is calculated. The combined white balance coefficients G _R , G _G , and G _B are calculated by the following Expression 13.

Equation 13 corrects the normal white balance coefficients G _Rw , G _Gw , and G _Bw with the skin color white balance coefficients G _Rf , G _Gf , and G _Bf . The degree of correction (weight) by the skin color white balance coefficients G _Rf , G _Gf , and G _Bf is determined by the constant α. The constant α may be a fixed value, and can be changed according to the reliability of the normal white balance coefficients G _Rw , G _Gw , G _Bw , and the flesh color white balance coefficients G _Rf , G _Gf , G _Bf. Also good.

For example, in the light source estimation process S205, when the divided areas are concentrated and distributed in the specific white area in FIG. 4, it is considered that the light source estimation accuracy is high. In this case, the reliability of the normal white balance coefficients G _Rw , G _Gw , and G _Bw is high, and it is considered that excessive correction with the skin color white balance coefficient is unnecessary, and the constant α is set to a small value.

  In addition, when the ratio of the person's face to the entire subject image is small, the number of divided areas to which a label indicating the skin color is added decreases, and the dispersion of independent components increases. In this case, it is considered that the reliability of the flesh color white balance coefficient is low, and the constant α is set to a small value.

Furthermore, in the light source estimation process S205, when the divided areas are distributed over the plurality of white areas in FIG. 4 and there is no significant difference in the number of plots between the white areas, the light source estimation accuracy is considered to be low. In this case, the reliability of the normal white balance coefficients G _Rw , G _Gw , and G _Bw is low, and it is considered that correction by the skin color white balance coefficients G _Rf , G _Gf , and G _Bf is necessary, and the constant α is set to a large value. Is done.

  Further, the method for calculating the composite white balance coefficient is not limited to Equation 13. For example, the combined white balance coefficient may be a sum of a white white balance coefficient and a skin color white balance coefficient in a predetermined ratio.

ホワイトバランス調整処理Ｓ１０４によって調整された撮像信号Ｒ_Ｐ´、Ｇ_Ｐ´、Ｂ_Ｐ´は、所定の画像処理が施され、ＬＣＤ１２４へのスルー画の表示や、撮影データの生成に使用される。 [S104 in FIG. 2 (White Balance Adjustment Processing)]
In this processing step S104, white balance adjustment processing is performed on the imaging signals R _P , G _P , and B _{P according} to the following formula 14.

The imaging signals R _P ′, G _P ′, and B _P ′ adjusted by the white balance adjustment processing S104 are subjected to predetermined image processing, and are used for displaying a through image on the LCD 124 and generating shooting data.

Thus, according to the present embodiment, independent components are extracted from the imaging signals R _P , G _P , and B _P based on the color information (color evaluation value) of the face area of the subject image. The independent components include a melanin pigment component M [i] and a hemoglobin pigment component H [i] that contribute to the color of human skin. Based on the statistics (distribution of independent components, skewness, weighted average, etc.) that characterize the distribution of these independent components M [i] and H [i], facial region feature amounts biasM and biasH are calculated. Skin color white balance coefficients G _Rf , G _Gf , and G _Bf are determined based on the amounts biasM and biasH. By performing the white balance adjustment process for the image pickup signal using the skin color white balance coefficients G _Rf , G _Gf , and G _Bf , it is possible to suitably reproduce the human skin color in the captured image.

  In addition, according to the present embodiment, for example, even when a subject image other than human skin is included in a part of the region of the subject image used for calculation of the skin color white balance coefficient, the statistics of the independent component By calculating the feature amounts biasM and biasH using, the influence of the subject image other than the person's skin on the skin color white balance coefficient can be suppressed, and appropriate white balance adjustment can be performed. Further, according to the present embodiment, the independent components M [i] and H [i] do not need to be sorted (rearranged) in accordance with the size of the values, so that the processing load on the image processing engine 116 is reduced. Appropriate white balance adjustment can be performed while suppressing.

  The above is the description of the exemplary embodiments of the present invention. Embodiments of the present invention are not limited to those described above, and various modifications are possible within the scope of the technical idea of the present invention. For example, the embodiment of the present application also includes an embodiment that is exemplarily specified in the specification or a combination of obvious embodiments and the like as appropriate.

  For example, in the coefficient determination process S307 illustrated in FIG. 5, the calculation formulas for the feature amounts biasM and biasH are not limited to Formula 10. The calculation method of the feature amounts biasM and biasH may be changed according to the number of divided regions to which labels are added.

  FIG. 10 is a plot of the independent components of each divided region when the number of divided regions to which labels are added is small. The horizontal axis of FIG. 10 indicates the size of the independent component M [i] corresponding to melanin, and the vertical axis indicates the size of the independent component H [i] corresponding to hemoglobin. FIG. 10 also shows a graph 10A showing the frequency of occurrence of the independent component M [i] and a graph 10B showing the frequency of occurrence of the independent component H [i]. The horizontal axis of the graph 10A indicates the value m [j] of the independent component M [i], and the vertical axis indicates the occurrence frequency frq_m. The horizontal axis of the graph 10B indicates the value h [k] of the independent component H [i], and the vertical axis indicates the occurrence frequency frq_h.

  As shown in graphs 10A and 10 in FIG. 10, the distribution of independent components when the number of divided regions is small is narrower than the graphs 9A and 9B in FIG. Also, as shown in the graph 10A, a value m [l] that never occurs may occur in the distribution of independent components. In this case, the reliability of independent component dispersion (divM, divH) and skewness (skwM, skwH) is low, and there is a possibility that appropriate feature amounts biasM, biasH cannot be obtained.

Therefore, when the number of divided regions to which labels are added is small, for example, the feature values biasM and biasH are set to the average values avgM and avgH of the independent component values, respectively, so that they do not greatly deviate from appropriate values. May be. In this case, the values of the reference feature amounts biasM ₀ and biasH ₀ are also changed according to the calculation method of the feature amounts biasM and biasH.

  If the feature amounts biasM and biasH are average values avgM and avgH, the contribution of the independent component having a small value to the feature amount becomes small, and the correlation between the color of the light source and the feature amount becomes low. However, it is possible to prevent the feature amount from deviating from an appropriate value due to low reliability of dispersion or distortion. The feature amounts biasM and biasH may be the median value or the mode value of the independent component values.

DESCRIPTION OF SYMBOLS 1 Image pick-up apparatus 100 System controller 102 Operation part 104 Drive circuit 106 Shooting lens 108 Aperture 110 Shutter 112 Image sensor 114 Image sensor drive circuit 116 Image processing engine 118 Buffer memory 120 Card interface 122 LCD control circuit 124 LCD
126 ROM
200 memory card

Claims (10)

An imaging signal storage unit that stores an imaging signal generated based on the luminous flux from the subject;
A skin detection unit for detecting a skin region in the subject image based on the imaging signal;
An independent component extraction unit that extracts an independent component from the color evaluation value of the skin region of the subject image;
A statistic calculator that calculates a statistic characterizing the distribution of the independent components;
A feature amount calculator that calculates a feature amount of the skin region based on the statistics,
A first coefficient calculation unit that calculates a first coefficient used for white balance adjustment of the imaging signal using the feature amount;
Comprising
White balance adjustment device. The feature amount calculation unit includes:
Set a reference value for the feature value, detect a difference between the reference value and the feature value,
Change the calculation method of the feature value and the reference value according to the reliability of the statistic,
The first coefficient calculation unit includes:
Calculating the first coefficient based on the difference detected by the feature amount calculation unit;
The white balance adjustment device according to claim 1. The statistic includes at least one of distribution of a plurality of independent component distributions, skewness of a plurality of independent component distributions, and a weighted average of the plurality of independent components.
The white balance adjustment apparatus according to claim 1 or 2. The statistic calculator is
Calculating a representative value of a plurality of independent component values;
The statistic is calculated by excluding an independent component whose value is more than a predetermined value from the representative value from a plurality of the independent components,
The white balance adjustment apparatus according to any one of claims 1 to 3. The independent component includes a melanin pigment component and a hemoglobin pigment component,
The white balance adjustment apparatus according to any one of claims 1 to 4. A second coefficient calculation unit that calculates a second coefficient used for white balance adjustment of the imaging signal based on a color evaluation value of the imaging signal by a predetermined algorithm;
A coefficient synthesizer that calculates a third coefficient used for white balance adjustment of the imaging signal based on the first coefficient and the second coefficient;
Further comprising
The white balance adjusting device according to any one of claims 1 to 5. An achromatic color detecting unit for extracting a region near the achromatic color in the subject image based on the imaging signal;
The second coefficient calculating unit calculates the second coefficient based on a color evaluation value of a region near the achromatic color;
The white balance adjusting device according to claim 6. An imaging signal storing step for storing an imaging signal generated based on a luminous flux from a subject;
A skin detection step of detecting a skin region in the subject image based on the imaging signal;
An independent component extraction step of extracting an independent component from the color evaluation value of the skin region of the subject image;
A statistic calculating step for calculating a statistic characterizing the distribution of the independent components;
A feature amount calculating step of calculating a feature amount of the skin region based on the statistics,
A first coefficient calculating step for calculating a first coefficient used for white balance adjustment of the imaging signal using the feature amount;
including,
White balance adjustment method. An imaging signal storing step for storing an imaging signal generated based on a luminous flux from a subject;
A skin detection step of detecting a skin region in the subject image based on the imaging signal;
An independent component extraction step of extracting an independent component from the color evaluation value of the skin region of the subject image;
A statistic calculating step for calculating a statistic characterizing the distribution of the independent components;
A feature amount calculating step of calculating a feature amount of the skin region based on the statistics,
A first coefficient calculating step for calculating a first coefficient used for white balance adjustment of the imaging signal using the feature amount;
A white balance adjustment program for causing a computer to execute a method including: An imaging unit that generates an imaging signal based on a luminous flux from a subject;
A skin detection unit for detecting a skin region in the subject image based on the imaging signal;
An independent component extraction unit that extracts an independent component from the color evaluation value of the skin region of the subject image;
A statistic calculator that calculates a statistic characterizing the distribution of the independent components;
A feature amount calculator that calculates a feature amount of the skin region based on the statistics,
A first coefficient calculation unit that calculates a first coefficient used for white balance adjustment of the imaging signal using the feature amount;
Comprising
Shooting device.

Images